AMD最新テクノロジーアップデート －HPCへの取り組み－ 日本AMD株式会社 マーケティング＆ビジネス開発本部 エンタープライズプロダクトマーケティング部 部長 山野 洋幸 AMD’s HPC Product Portfolio Energy efficient CPU and discrete GPU processors focused on addressing the most demanding HPC workloads
Multi-core x86 Processors • Outstanding Performance • Superior Scalability • Enhanced Power Efficiency
Professional Graphics • 3D Accelerators For Visualization • See More and Do More with Your Data
ATI Stream Computing • GPU Optimized For Computation • Massive Data-parallel Processing • High Performance Per Watt
2 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD’s HPC Product Portfolio Energy efficient CPU and discrete GPU processors focused on addressing the most demanding HPC workloads
Multi-core x86 Processors • Outstanding Performance • Superior Scalability • Enhanced Power Efficiency
Professional Graphics • 3D Accelerators For Visualization • See More and Do More with Your Data
ATI Stream Computing • GPU Optimized For Computation • Massive Data-parallel Processing • High Performance Per Watt
3 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Planned Server Platform Roadmap 2006 2007 2008 2009 2010 2011 “Maranello” Socket G34 with AMD SR56x0 and SP5100 Magny-Cours New Architecture Six-Core AMD Opteron™ Processor with AMD
way Chipset - Socket F(1207) with AMD SR56x0 and SP5100
Shanghai/Istanbul
Platform 2/4 Enterprise Enterprise “Socket F (1207)” Socket F(1207) with Nvidia and Broadcom chipsets Rev F Barcelona Shanghai Istanbul “San Marino” Socket C32+AMD SR56x0 and SP5100
Lisbon New Architecture way
- “Adelaide” Platform 1/2 Socket C32 EE+AMD SR5650 and SP5100
Lisbon New Architecture Power Efficient Efficient Power “Buenos Aires” Socket AM3 with AMD SR56x0 and SP5100
Suzuka
way -
1 “Socket AM2+” Platform Socket AM2+ with Nvidia and Broadcom chipsets 4 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 Rev F Budapest Suzuka For more information be sure to visit AMD at SC’09 booth #1417 Planned Server Platform Roadmap 2006 2007 2008 2009 2010 2011 “Maranello” Socket G34 with AMD SR56x0 and SP5100 Magny-Cours New Architecture Six-Core AMD Opteron™ Processor with AMD
way Chipset - Socket F(1207) with AMD SR56x0 and SP5100
Shanghai/Istanbul
Platform 2/4 Enterprise Enterprise “Socket F (1207)” Socket F(1207) with Nvidia and Broadcom chipsets Rev F Barcelona Shanghai Istanbul “San Marino” Socket C32+AMD SR56x0 and SP5100
Lisbon New Architecture way
- “Adelaide” Platform 1/2 Socket C32 EE+AMD SR5650 and SP5100
Lisbon New Architecture Power Efficient Efficient Power “Buenos Aires” Socket AM3 with AMD SR56x0 and SP5100
Suzuka
way -
1 “Socket AM2+” Platform Socket AM2+ with Nvidia and Broadcom chipsets 5 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 Rev F Budapest Suzuka For more information be sure to visit AMD at SC’09 booth #1417 Six-Core AMD Opteron™ Processor
• Six true cores • New HyperTransport™ technology HT Assist • Increased HyperTransport 3 technology (HT3) bandwidth • Higher performing Integrated Memory Controller • Same power/thermal ranges as Quad-Core AMD Opteron™ processor • Up to 50% higher performance than Quad-Core AMD Opteron™ processor-based servers at the same processor ACP1
Model Cores Freq NB Wattage L2 Cache L3 Cache Production 8435 2.6GHz 8431 2.4GHz 2435 6 2.6GHz 2.2GHz 75W 512K/core 6MB May 2431 2.4GHz 2427 2.2GHz 8425 HE 2.1 GHz 2425 HE 6 2.1 GHz 2.2 GHz 55W 512K/core 6MB July 3rd 2423 HE 2.0 GHz 8439 SE 2.8 GHz 6 2.2 GHz 105W 512K/core 6MB July 24th 2439 SE 2.8 GHz 2419EE 6 1.8GHz 2.0GHz 40W 512K/core 6MB September
6 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 *Based on SPECint® 2006 results published as of 11/09/09. See backup slides or configuration and performance information. For more information be sure to visit AMD at SC’09 booth #1417 Six-Core AMD RDDR-2 RDDR-2 Opteron™ Processor Up to 8 DIMMs Up to 8 DIMMs
HT-1
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM Performance “Istanbul” HT-3 “Istanbul” DIMM • Six-Core AMD Opteron™ Processor 6M Shared L3 Cache
North Bridge enhancements (PF + prefetch)
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM 45nm Process Technology DIMM • DDR2-800 Memory HT-1 HT-1 • HyperTransport-3 @ 4.8 GT/sec HT-3 HT-3
HT-1
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM Reliability/Availability “Istanbul” HT-3 “Istanbul” DIMM • L3 Cache Index Disable
• HyperTransport Retry (HT-3 Mode)
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM • x8 ECC (Supports x4 Chipkill in unganged mode) DIMM HT-1 HT-1 Virtualization HT-3 HT-3 • AMD-V™ with Rapid Virtualization Indexing PCI-e PCI-e or I/O Bridge or Manageability PCI-X PCI-X • APML Management Link*
Scalability USB • 48-bit Physical Addressing (256TB) South SATA SE: 2.8GHz • HT Assist (Cache Probe Filter) PCI Bridge Std: 2.6GHz LPC PATA HE: 2.1GHz Continued Platform Compatibility EE: 1.8GHz • Nvidia/Broadcom-based F/1207 platforms 4 socket example block diagram
*APML-enabled platform support required. 7 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Six-Core AMD Opteron™ RDDR-2 RDDR-2 Processor Up to 8 DIMMs Up to 8 DIMMs
Performance HT-1
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM “Istanbul” “Istanbul” DIMM • Six-Core AMD Opteron™ Processor HT-3 6M Shared L3 Cache
North Bridge enhancements (PF + prefetch)
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM 45nm Process Technology DIMM • DDR2-800 Memory HT-1 HT-1 • HyperTransport-3 @ 4.8 GT/sec HT-3 HT-3
HT-1
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM Reliability/Availability “Istanbul” HT-3 “Istanbul” DIMM • L3 Cache Index Disable
• HyperTransport Retry (HT-3 Mode)
DIMM DIMM DIMM DIMM
DIMM DIMM DIMM • x8 ECC (Supports x4 Chipkill in unganged mode) DIMM HT-1 HT-1 Virtualization HT-3 HT-3 • AMD-V™ with Rapid Virtualization Indexing 4 socket example block diagram
Manageability STREAM Bandwidth • APML Management Link* (GB/s)* Scalability Product, Freq, Dram 2S 4S 8S • 48-bit Physical Addressing (256TB) • HT Assist (Cache Probe Filter) “Barcelona,” 2.3/2.0, RDDR2-667 17.2 20.5 “Shanghai,” 2.7/2.2, RDDR2-800 21.4 24 Continued Platform Compatibility • Nvidia/Broadcom-based F/1207 platforms “Istanbul,” 2.4/2.2, RDDR2-800 22 42 81.5
*Based on measurements at AMD performance labs. See backup slide for configuration information. 8 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HT Assist Improves Two-Socket to Four-Socket Server Performance Scaling
200% 200% STREAM (GB/s) STREAM (GB/s) +100% +93% +96% 180% SPECfp®_rate2006 180% SPECfp®_rate2006 +82% SPECint®_rate2006 +80% SPECint®_rate2006 160% 160% Up to 82% higher Up to 2x 140% 140% performance performance 120% 120% +19% 100% 100%
80% 80%
RelativePerformance RelativePerformance 60% 60%
40% 40%
20% 20%
0% 0% Quad-Core AMD Opteron™ processor Quad-Core AMD Opteron™ processor Six-Core AMD Opteron™ processor Six-Core AMD Opteron™ processor ("Shanghai") ("Shanghai") ("Istanbul") ("Istanbul") Model 2384 (2.7GHz) (2P) Model 8384 (2.7GHz) (4P) Model 2435 (2.6GHz) (2P) Model 8435 (2.6GHz) (4P)
SPEC, SPECint, and SPECfp are registered trademarks of the Standard Performance Evaluation Corporation. The SPECint_rate and SPECfp_rate results for Six-Core AMD OpteronTM processors are based upon data submitted to Standard Performance Evaluation Corporation as of May 12, 2009. The other SPECint_rate and SPECfp_rate results stated above reflect results published on http://www.spec.org/ as of May 12, 2009. The STREAM results are based on measurements at AMD performance labs as of May 12, 2009. The SPECint_rate and SPECfp_rate comparisons presented above are based on the best performing two-socket and four-socket servers using AMD Opteron™ processor Models 2384, 8384, 2435, and 8435. For the latest results, visit www.spec.org. Please see backup slides for configuration information.
9 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 2 Socket HPL Performance High-Performance Linpack - Two Socket Servers
Six-Core AMD Opteron™ 104 Processor ("Istanbul") (139%) Model 2435 (2.6GHz)
Intel Xeon Processor 75 ("Gainestown") Model (100%) X5570 (2.93GHz)
0 20 40 60 80 100 120 HPL Gflops*
*Based on AMD internal performance lab measurements as of June 15th 2009 – see backup slides for configuration
10 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Power Efficient Innovations
AMD PowerNow!™ Technology with Independent Dynamic Core Technology Allows processors and cores to dynamically operate at lower power AMD Smart Fetch and frequencies, depending on usage and workload to help reduce Technology TCO and to lower power consumption in the datacenter Can reduce power consumption by allowing idle cores to enter a “halt” state Enhanced Performance-per-watt 50% more compute cores vs. quad-core within the same power envelope*
AMD PowerCap AMD CoolCore™ Manager Technology Allows IT datacenter Can reduce processor managers to set a energy consumption by fixed limit on a dynamically turning off server's processor sections of the processor power consumption when inactive
Extends to the L3 Cache
Dual Dynamic Power Management™ Enables more granular power management capabilities to reduce processor energy consumption. Separate power planes for cores and memory controller
11 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 *Compared to Quad-Core AMD Opteron processor codenamed “Shanghai.” For more information be sure to visit AMD at SC’09 booth #1417 Significantly Higher Performance in Same Power Envelope
Two-socket servers using Six-Core AMD Opteron™ processors significantly outperform two-socket servers using Dual-Core and Quad-Core AMD Opteron™ processors without consuming significantly more power
700% SPECfp®_rate2006 600% SPECint®_rate2006 SPECjbb®2005 500% Server Idle Power
400% 3x core count >3x performance 300% Same power envelope
200% Relative Performance Relative
100%
0% Dual-Core Quad-Core Six-Core AMD Opteron™ processor AMD Opteron™ processor ("Shanghai") AMD Opteron™ processor ("Istanbul") Model 2218 (2.6GHz) Model 2382 (2.6GHz) Model 2435 (2.6GHz)
SPEC, SPECfp, SPECint, and SPECjbb are registered trademarks of the Standard Performance Evaluation Corporation. The performance results for Six-Core AMD Opteron™ processor Model 2435 and the SPECjbb result for Quad-Core AMD Opteron™ processor Model 2382 are based upon data submitted to Standard Performance Evaluation Corporation as of May 21, 2009. The other performance results stated below reflect results published on http://www.spec.org/ as of May 21, 2009. The server idle power results are based on measurements of server active idle power for 60 seconds at AMD performance labs as of May 21, 2009. The comparison presented below is based on the best performing two-socket servers using AMD Opteron™ processor Models 2218, 2382, and 2435. For the latest results, visit http://www.spec.org/. Please see backup slides for configuration information.
12 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Six-Core AMD Opteron™ EE Processor
- Only Six-Core 40W ACP for 2P servers
- Targeted at energy-efficient servers and workloads such as cloud computing, Web 2.0 applications, and dense computing environments
- Low power without compromise; Built with AMD Virtualization™ (AMD-V™) technology and AMD-P
- Up to 31% higher performance-per-watt than Quad Core AMD Opteron™ 75W processor*
- Up to 33% higher performance over previous Quad-Core AMD Opteron™ EE processor**
*Compared to 75W ACP Quad-Core AMD Opteron™ processor. See backup slides for configuration information. **Compared to Quad-Core AMD Opteron™ EE processor. See backup slides for configuration information.
13 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Low Power without compromise
Six-Core AMD Opteron™ Six-Core AMD
HE Opteron™ EE
100%
75%
50%
™ ™ technology speed
25%
P P
V™ V™
- -
- -
AMD AMD
Maximum Memory Speed technology speed HyperTransport™
Maximum Memory Speed HyperTransport
AMD AMD
Cache Cache
Consistent Feature Set Inconsistent Feature Set
Intel specifications as of 5/15/09: http://www.intel.com/Assets/PDF/prodbrief/xeon-5500.pdf 14 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Launched: Six-Core AMD Opteron™ Processor with AMD Chipset System architecture
Unified server platforms with AMD on AMD – benefits of processor + chipset Enhanced platforms with chipset choices and key technologies . Ultra power efficient platform designed for cloud and dense deployments (using AMD Opteron™ EE processor and S5650 chipset) . First 4P x86 Platform with PCI Express® 2.0 for HPC SR56x0 SR56x0 I/O Hub I/O Hub applications AMD-V™ AMD-V™ . Three chipset solutions to optimize workloads Extends AMD Virtualization™ (AMD-V™) technology and SP5100 South AMD-P suite of features Bridge Wide availability from ODMs; OEMs are expected to introduce new chipsets as part of AMD’s 2010 processor lineup
15 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Benefits of Six-Core AMD Opteron™ processor with AMD Chipset
Choice of Consistent Key Chipsets Feature Set Technologies
® . AMD-P . PCI Express 2.0 SR5690 SR5670 SR5650 . AMD-V™ . HyperTransport™ 3.0 Technology
Benefits
Range and flexibility Exceptional of chipsets Virtualization Increased (from high bandwidth to capabilities performance and power efficient) throughput (up to 2x to meet the demands Only 4P x86 solution for PCI Express® with Tagged TLB, RVI of your workloads 2.0 vs. PCI and I/O Virtualization Express® 1.0) Power efficiency beyond the processor
16 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HPC Scalability - Think Outside The Box Scale Up: More processors, more cores & improved scalability (HT Assist) Scale Out: Multiple servers networked together into a large powerful computer
AMD’s new SR56x0 chipset helps ensure platform level performance innovations can be realized at scale
HyperTransport™ 3.0 technology and PCI-E® 2.0 provide the low latency and high bandwidth necessary to take full advantage of high performance interconnects such as QDR InfiniBand
17 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HPC Scalability - Think Outside The Box
AMD SR56x0 chipset with QDR IB delivers up to 22% reduced latency and 89% increased bandwidth over prior generation chipsets with DDR IB*
Unidirectional Bidirectional IB test Latency (usec) Bandwidth Bandwidth (MB/sec) (MB/sec) Affinity C0-c0 C4-c4 C0-c0 C4-c4 C0-c0 C4-c4 PCIe® Gen1 + 1.17 1.21 1549 1542 2918 2900 HT1 + IB DDR PCIe® Gen2 + 0.91 0.96 2773 2767 5486 5495 HT3 + IB QDR % diff -22% -20% 79% 79% 88% 89%
* Based on AMD internal performance lab measurements 18 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HPC University Implementation with PRACE Challenge Solution
Collaborative effort between Supermicro UK office Customer: Swedish National Infrastructure for and AMD Computing (SNIC), Royal Institute of Technology AMD product: Six-Core AMD Opteron™ HE (KTH), Sweden, jointly with the Partnership for processor with new AMD SR5670 chipset using Advanced Computing in Europe (PRACE) HyperTransport™ 3 technology and HT assist
Need: A general purpose HPC cluster for energy Infrastructure: efficiency and compute density using standard . Supermicro Highly dense 7U enclosure components (with 10x 4P Blade Model) BHQIE . QDR Switch incorporated inside the Buying Criteria: blade enclosure . Superior performance/watt/density and ability . Management per node including to control energy consumption depending on KVM/IP workload and performance based on flop/s and . Redundant PSU’s capable of operating GB/s at ultra high power efficiency . x86 Platform . Latest technology chipset supporting PCI Quantity: 720 processors; 4320 cores Express® 2.0
19 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD Developer Central – developer.amd.com // Code Faster, Faster Code
“Istanbul” Zone What’s in it for software developers? - Find out at developer.amd.com
Visit the developer site regularly for updated information and practical guidance on how to take advantage of all the new features in the Quad- Core AMD Opteron™ processor, and Six-Core AMD Opteron™ processor.
» Software Development Tools and Resources » Overview of Software Visible Features » Documentation » Technical Articles & Blogs » Benchmarks and Performance Evaluations
20 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD support for GCC
http://developer.amd.com/cpu/gnu/Pages/default.aspx
21 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 New! - ACML 4.3
. L3 BLAS improvements – SGEMM for Six-Core AMD Opteron™ processor – New Intel DGEMM and SGEMM kernels . Supporting Woodcrest, Penryn, Nehalem – New DGEMM "fast memory allocation" scheme . allows improved performance of other routines (such as LAPACK) which make heavy use of DGEMM . Six-Core AMD Opteron™ processor tuning for Level 1 BLAS – xDOT, xCOPY, xAXPY, and xSCAL . 3DFFT performance improvements . AMD Family 10h tuning for real-complex FFTs – csfft, dzfft, scfft and zdfft have been re-tuned for AMD Family 10h processors, providing significant performance increases.
22 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 New - AMD’s High Performance Compiler
AMD has released a version of C/C++/Fortran compilers based on Open64 technology – x86 Open64 version 4.2.2.2 . Release is available for download on AMD Developer Central at http://developer.amd.com/cpu/open64 in source and binary forms . No license fee for compiler use . Provided with documentation and support
File Name Launch Date OS Bitness Description x86_open64-4.2.2.2- Source code for x86 08/31/2009 Linux® Host 32/64-bit 1.src.tar.bz2 (59.6MB) Open64 4.2.2.2-1 x86_open64-4.2.2.2- Host 64-bit x86 Open64 4.2.2.2-1 08/31/2009 RHEL 5.3 & SLES 10 SP2 1.x86_64.rpm (23.4MB) Target 32/64-bit Compilers for Linux x86_open64-4.2.2.2- Host 64-bit x86 Open64 4.2.2.2-1 08/31/2009 RHEL 5.3 & SLES 10 SP2 1.x86_64.tar.bz2 (23.4MB) Target 32/64-bit Compilers for Linux
23 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Software Optimization & Compiler Guidelines
http://developer.amd.com/documentation/guides/Pages/default.aspx
24 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 x86 64-bit Architecture Evolution
2003 2005 2007 2008 2009 2010*
AMD AMD “Magny- “Barcelona” “Shanghai” “Istanbul” Opteron™ Opteron™ Cours”
Mfg. 90nm SOI 90nm SOI 65nm SOI 45nm SOI 45nm SOI 45nm SOI Process K8 K8 “Greyhound” “Greyhound+ “Greyhound+” “Greyhound+ CPU Core ”
L2/L3 1MB/0 1MB/0 512kB/2MB 512kB/6MB 512kB/6MB 512kB/12MB
Hyper Transport™ 3x 1.6GT/.s 3x 1.6GT/.s 3x 2GT/s 3x 4.0GT/s 3x 4.8GT/s 4x 6.4GT/s Technology
Memory 2x DDR1 300 2x DDR1 400 2x DDR2 667 2x DDR2 800 2x DDR2 800 4x DDR3 1333
*Planned model or feature introduction dates.
25 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD 2010-2011 Sweet Spot Server Strategy
2010 2011
4P/8P Platforms ~5% of Market* “Magny-Cours” “Interlagos” AMD Opteron™ 8 and 12 cores 12 and 16 cores 6000 Series Platform Performance-per- ”Maranello” • 2/4 socket; 4 memory channels watt and • Highly scalable without compromising value Expandability
2P Platforms ~75% of Market* Platform Consistency Bulldozer and Commonality Core
Highly Energy AMD Opteron™ “Lisbon” “Valencia” Efficient and 4000 Series Platform 4 and 6 cores 6 and 8 cores “San Marino” Cost Optimized • 1/2 socket; 2 memory channels 1P Platforms and “Adelaide” • New levels of value and power efficiency ~20% of Market*
26 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 *AMD internal estimates of total server market as of Q309 For more information be sure to visit AMD at SC’09 booth #1417 Unmatched Feature Consistency Across Platforms
AMD Opteron™ 6000 Series Platform Same BIOS (G34 Socket)
Shared Platform Interface Same API
AMD Opteron™ 4000 Series Platform (C32 Socket)
Same Drivers
. OEM Benefits: enables easier design, validation and support . Customer Benefits: easy management and flexibility
27 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 The AMD Opteron™ 6100 Processor
. Target: Enterprise Class 2-way and 4P Socket G34 Server Platform 4-way Servers 12/8 core Processor Support – Twelve-core and Eight–core 12M L3 Cache – CoolCore™ Technology, Enhanced AMD PowerNow!™ Technology, Enhanced C1 state, AMD CoolSpeed™ Technology, APML – Quad-Channel LV & U/RDDR3, ECC, On-line spare – Up to 3 DIMMs/channel, 12 per CPU Coherent HT3 – Expected platforms 2P/2U, 2P Tower, 4P rack, 4P Blade . Single Series for performance DP and MP platforms Non-coherent HT3 – 2P economics for 4P servers – Compelling price/performance for volume SR56x0 SR56x0 market AMD-Vi™ AMD-Vi . G34 Socket Infrastructure x4 A-Link – Performance-optimized Power/thermals
– Quad 16-bit HT3 links, up to 6.4 GT/s per SP5100 link South Bridge – AMD SR56x0 chipset with AMD-Vi™ and PCIe Gen2
28 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HyperTransport™ Technology HT Assist (Probe Filter)
Key enabling technology on Six-Core AMD “Old” broadcast protocol Opteron™ processor and AMD Opteron™ 6100 Home Series processor Node Probes
RdBlk HT Assist is a sparse directory cache Request Resps Req . Associated with the memory controller on Node the home node . Tracks all lines cached in the system from PF – clean data Home PF the home node Node Lookup
DRAM Eliminates most probe broadcasts (see diagram) Resp Req . Lowers latency Node – local accesses get local DRAM latency, no need to wait for probe responses PF – dirty data Home PF – less queuing delay due to lower HT Node Lookup traffic overhead Directed Probe . Increases system bandwidth by reducing Req Cache Owner probe traffic Node Resp
29 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HT Assist and Memory Latency
With “old” broadcast coherence protocol, the latency of a memory access is the longer of 2 paths: . time it takes to return data from DRAM and . the time it takes to probe all caches
With HT Assist, local memory latency is significantly reduced as it is not necessary to probe caches on other nodes.
Several server workloads naturally have ~100% local accesses . SPECint®, SPECfp® . VMMARK typically run with 1 VM per core . SPECpower_ssj with 1 JVM per core . STREAM
Probe Filter amplifies benefit of any NUMA optimizations in OS/application which make memory accesses local
SPEC, SPECint, SPECfp, and SPECpower_ssj are trademarks or registered trademarks of the Standard Performance Evaluation Corporation. 30 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Enhanced Integrated Memory Architecture Benefits HPC, Virtualization and Database
Memory Bandwidth Memory Bandwidth (GB/s in STREAM benchmark) (GB/s in STREAM benchmark)
60 4P AMD Opteron™ 61XX 50 8P AMD Opteron 84XX 40 4P AMD Opteron 84XX 30 4/8 Socket 4/8 4P Intel Xeon 74XX 20
10 2P AMD Opteron 61XX 0 2P Intel Xeon 55XX High Power Standard Power Low Power
2P AMD Opteron 24XX 2P AMD Opteron 6100 2P Intel Xeon 55XX 2 Socket 2
0 50 100 …and consistency across Greater peak performance… power bands
1) Based on measurements in AMD Performance Labs as of October 6, 2009. Please see backup slides for configuration information.
31 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Customer Support for AMD Strategy
“We’ve been able to provide remarkable value to our customers, in part due to AMD’s platform consistency and scale up/scale out capability. Our HPC customers can quickly and productively utilize all the processor performance we can provide and they are looking forward to the next wave of innovation with the 8- and 12-core “Maranello” platform. For our part, we are very happy with what we’re seeing in the early silicon and plan to have systems available in 1H10.” Barry Bolding, Vice President, Scalable Systems
32 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Server Roadmap 65nm 45nm 32nm Yellow text denotes new feature
“Shanghai” “Istanbul” “Magny-Cours” “Interlagos” 4-Core 6-Core 8/12-Core 12/16-Core • 6M L3 • 6M L3 • 12M L3 New Core • 3x HT-3 (4.4GT) • 3x HT-3 (4.8GT) • 4x HT-3 (6.4GT) • AMD-V technology • HT Assist • U/RDDR3 & LV RDDR3 (Quad-Channel) • RDDR2 (Dual-Channel) • AMD-V technology • Cool Speed • RDDR2 (Dual-Channel) • C1E 4-way • AMD-V 2 and 4-way Performance • HT Assist Enterprise/ Platform Mainstream “Socket F (1207)” Platform “Maranello” “Six-Core AMD Opteron™ Processor w/AMD Chipset” • AMD SR56x0 ● AMD SP5100 “Maximum Scalability” ● APML Enabled (Istanbul Only) • AMD SR56x0 • AMD SP5100 • Advanced Platform Management “Shanghai” “Istanbul” 4-Core 6-Core
2-way Mainstream “Lisbon” “Valencia” 4/6-Core Platform 6/8-Core “Socket F (1207)” • 6M L3 New Core • 2x HT-3 (6.4GT) • U/RDDR3 & LV RDDR3 (Dual-Channel) “Six-Core AMD Opteron™ Processor with AMD Chipset” • Cool Speed • C1E • HT Assist 1 and 2-way • AMD-V “Suzuka” • 6M L3 Energy “Budapest” • DDR3 Efficient/ 4-Core • 1xHT3 4-Core • AMD-V technology Cost Optimized Platform • AMD SR56x0 “San Marino” (Std/HE/EE) 1-way “Socket AM2+” • AMD SP5100 Platform “Optimized Energy Efficiency” • Advanced Platform Management
“Buenos Aires” • AMD SR5650 “Adelaide” (EE Only) • AMD SP5100 • AMD SR56x0 • LV DDR3 • AMD SP5100 “Ultra Low Power” • HT1
Platform Segment 2009 2010* 2011*
33 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 * Planned roadmaps For more information be sure to visit AMD at SC’09 booth #1417 Introducing the “Bulldozer” Architecture
. “Bulldozer” is an innovative Fetch new architecture that dramatically Decode reduces the thread- Int Int Scheduler interference Scheduler
. Only AMD is planning to FP Scheduler deliver the most scalable
industry-standard bit
processor architecture with bit
- -
Pipeline Pipeline Pipeline Pipeline Pipeline Pipeline Pipeline
true core functionality Pipeline
FAMC FAMC
128 128
. “Bulldozer” is a modular L1 DCache L1 DCache
architecture that creates the Shared L2 Cache building blocks of the next generation of processor Shared L3 Cache and NB designs
34 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Designed for Scalability and Performance
“Bulldozer” Flex FP module
Two cores in a single unit that enables two A flexible floating point simultaneous threads, the unit that can be dedicated building blocks of a OR shared between the “Bulldozer” die two cores per cycle
Parallel Dedicated Threads Scheduler
The ability to execute two Independent integer threads on two discrete, schedulers and an FP unshared cores without scheduler improve compromising or creating scalability by efficient bottlenecks execution
35 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Unprecedented Server Performance Gains
45 Floating Point Integer 40
35
30
25
20
15
10
5
0 2003 2004 2005 2006 2007 2008 2009 2010 2011
AMD Opteron™ AMD Opteron AMD Opteron AMD Opteron AMD Opteron AMD Opteron AMD Opteron “Magny-Cours”* “Interlagos”* 244 250 x75 285 x356 x384 2435 (dual-core) (dual-core)
2010 is projected to be the beginning of unprecedented leaps in server performance-per-watt for AMD
36 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 * “Magny-Cours” and “Interlagos” data is based on AMD projections For more information be sure to visit AMD at SC’09 booth #1417 AMD’s HPC Product Portfolio Energy efficient CPU and discrete GPU processors focused on addressing the most demanding HPC workloads
Multi-core x86 Processors • Outstanding Performance • Superior Scalability • Enhanced Power Efficiency
Professional Graphics • 3D Accelerators For Visualization • See More and Do More with Your Data
ATI Stream Computing • GPU Optimized For Computation • Massive Data-parallel Processing • High Performance Per Watt
37 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Harnessing the Computational Power of GPUs
• GPU architecture increasingly emphasizes programmable shaders instead of fixed function logic • Enormous computational capability for data parallel workloads • Potential to balance system performance • New math for datacenters: high performance/watt and performance/$
Software Applications
Graphics Serial and Workloads Task Parallel Workloads Data Parallel Workloads
38 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ATI Stream Technology is… Heterogeneous: Developers leverage AMD GPUs and CPUs for optimal application performance and user experience
Industry Standards: OpenCL™ and DirectCompute 11 enable cross-platform development
High performance: Massively parallel, programmable GPU architecture delivers advanced performance and power efficiency
Gaming Digital Content Productivity Creation
Sciences Government Engineering
39 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 GPGPU Processing Power Trend
3000 * Cypress GigaFLOPS ATI RADEON™ HD 5870 2500 RV770 ATI RADEON™ HD 4800 RV670 ATI FirePro™ 2000 ATI RADEON™ V8700 R600 HD 3800 AMD FireStream™ ATI RADEON™ ATI FireGL™ 9250 OpenCL™ 1.0 HD 2900 V7700 9270 R580(+) ATI FireGL™ AMD FireStream™ DirectX® 11 ATI RADEON™ 9170 1500 X19xx V7600 R520 ATI FireStream™ V8600 ATI RADEON™ V8650 2.5x ALU X1800 ATI FireGL™ increase 1000 V7200 V7300 Stream SDK V7350 CAL+IL/Brook 500 + Double-precision GPGPU Unified floating point via CTM Shaders 0 support
Jul-09 Sep-05 Mar-06 Oct-06 Apr-07 Nov-07 Jun-08 Dec-08
*Chart indicated peak theoretical single-precision performance.
40 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Focus On Efficient GPU Progression
14.47 GFLOPS/W
GFLOPS/W GFLOPS/mm2 7.50
4.50 7.90 GFLOPS/mm2
2.01 2.21 4.56 1.07 2.24 0.42 1.06 0.92
41 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Enhancing GPUs for Computation AMD FireStream™ 9270 Stream Processor • 800 stream processing units arranged in 10 SIMD cores • Up to 1.2 TFLOPS peak single precision floating point performance; Fast double-precision processing w/ up to 240 GFLOPS • Up to 115 GB/sec GDDR5 memory interface • Up to 480 GB/s L1 & up to 384 GB/s L2 cache bandwidth • Data sharing between threads • Improved scatter/gather operations for improved GPGPU memory performance* • Integer bit shift operations for all units – useful for crypto, compression, video processing • More aggressive clock gating for improved performance per watt*
*Compared to previous generation AMD FireStream™ stream processors. 42 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009 42 For more information be sure to visit AMD at SC’09 booth #1417 ATI Radeon™ HD 5870 Graphics Architecture
2.72 Teraflops Single Precision 544 Gigaflops Double Precision
• Full Hardware Implementation of DirectCompute 11 and OpenCL™ 1.0
• IEEE754-2008 Compliance Enhancements
• Additional Compute Features: • 32-bit Atomic Operations • Flexible 32kB Local Data Shares • 64kB Global Data Share • Global synchronization • Append/consume buffers
43 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Harnessing the Computational Power of GPUs
Unleash the full potential of STREAM computing with AMD’s new SR56x0 chipsets
• GPU architecture increasingly emphasizes programmable shaders instead of fixed function logic • Enormous computational capability for data parallel workloads • New math for datacenters: high performance/watt and performance/$ • Open systems approach to enable developers: active in Khronos OpenCL™ Working Group driving new cross-platform API
44 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Moving Past Proprietary Solutions for Ease of Cross-Platform Programming
Open and Custom Tools High Level Tools High Level Language Application Specific Compilers Libraries
Industry Standard Interfaces
OpenCL™ DirectX® OpenGL®
AMD AMD Other GPUs CPUs CPUs/GPUs
• Cross-platform development • Interoperability with OpenGL and DirectX® • CPU/GPU back-ends enable balanced platform approach
45 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ATI Stream SDK v2.0 Beta: OpenCL™ For Multicore x86 CPUs and GPUs The Power of Fusion: Developers leverage heterogeneous architecture to deliver superior user experience • First complete OpenCL™ development platform • Certified OpenCL™ 1.0 compliant by the Khronos Group • Write code that can scale well on multi-core CPUs and GPUs • AMD delivers on the promise of OpenCL™, with both high- performance CPU and GPU technologies • Available for download now as part of ATI Stream SDK beta program – includes documentation, samples, and developer support Beta Program: http://developer.amd.com/streambeta
Software Applications
Graphics Workloads Serial and Data Parallel Task Parallel Workloads Workloads
OpenCL and the OpenCL logo are trademarks of Apple Inc. used by permission by Khronos. 46 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 First To Support DirectCompute 11
• First and only driver to support DirectCompute 11 with Shader Model 5.0. • Fully Microsoft® WHQL certified • Available with ATI Radeon™ HD 58xx Series
47 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ACML-GPU 1.0 Released March 2009 Selected BLAS routines enabled for GPU . DGEMM, SGEMM will run on GPU if present . Small problems (N,M,K < 200) run on CPU . Supports Quad- and Six-Core AMD Opteron™ processors BLAS LAPACK FFT RNG LIBMV Expert L3 23 Shape Array Simple Generators L2 LAPACK Vector 3.1.1 3D L1 2D 5 Base Fast Scalar 1D Generators
48 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD’s HPC Product Portfolio Energy efficient CPU and discrete GPU processors focused on addressing the most demanding HPC workloads
Multi-core x86 Processors • Outstanding Performance • Superior Scalability • Enhanced Power Efficiency
Professional Consumer Graphics • 3D Accelerators For Visualization • See More and Do More with Your Data
ATI Stream Computing • GPU Optimized For Computation • Massive Data-parallel Processing • High Performance Per Watt
49 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ATI Radeon™ HD 5870 GPU
The World’s Most Powerful and Advanced GPU
. Ultimate immersion with DirectX® 11 and ATI Eyefinity Technology . Accelerating PCs with nearly 3 teraFLOPS of compute power . Extreme game play at high resolutions and maximum settings**
Compute Power Up to 2.72 TFLOPS
Core Clock Speed 850 MHz
Stream Processors 1600
Frame Buffer 1GB , 4.8Gbps
Max/Idle Board Power 188W/27W
Max Res. 3x 2560x1600**
*Claim of most powerful processor based on raw compute horsepower measured in FLOPS. Internal calculations show that the processor used in the ATI Radeon HD 5800 series can achieve 2.72 TeraFLOPS, more than any other known microprocessor as of September 23, 2009. FLOPS is not necessarily an indicator of leading performance in every application as AMD GPUs are designed and built to excel specifically in applications involving massively parallel calculations. **Subject to monitor resolution
50 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ATI Eyefinity Technology Think of it as surround-sight Multiple displays are treated as a single display by the OS.*
Supports up to six 30” panels with a combined resolution of up to 7680 x 3200**
Powering the displays was a SINGLE graphics card, with a SINGLE GPU† *To enable a third display requires one panel with a DisplayPort connector or an active DisplayPort adapter. **The ATI Catalyst™ driver 8663.1 is required to support 3x 2560x1600 resolution. † With the introduction of AMD’s next‐generation graphics product series supporting DirectX® 11, a single GPU now has the advanced capability of simultaneously supporting up to six independent display outputs.
51 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 ATI Eyefinity – YouTube Videos
11,520 x 4,800 - 55.3 megapixels, Four Cards, single PC
http://www.youtube.com/watch?v=mzGtxlaPQqY http://www.youtube.com/watch?v=N6Vf8R_gOec
For more information:
http://www.amd.com/us/products/technologies/eyefinity/Pages/eyefinity.aspx http://www.amd.com/us/Documents/ATI_Eyefinity_Technology_Brief.pdf http://blogs.amd.com/home/2009/09/10/ati-eyefinity%E2%80%99s-panoramic-future-keep-watch/
52 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD at Supercomputing 2009
AMD will be in booth #1417 at Supercomputing 2009 in Portland, Oregon on November 16th through 19th.
Be one of the first to preview the upcoming 8 and 12 core AMD Opteron™ 6000 Series platforms featuring the AMD Opteron™ 6100 Series processors. This new platform provides an outstanding new option for the next generation of HPC innovation.
We will showcase solutions based on AMD FireStream stream processor, ATI FirePro™ workstation graphics accelerators, and the leading-edge AMD Opteron™ processors.
Visit our events page @ www.amd.com/supercomputing09 and learn more about our technology partners featuring AMD technology-based solutions at the SC09, and get information on exciting promotions.
53 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 THANK YOU!
54 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 BACKUP AND CONFIGURATION INFORMATION
55 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Where Do We Put the HT Assist Probe Filter?
Q: Where do we store probe filter entries without adding a large on-chip probe filter RAM which is not used in a 1P desktop system?
A: Steal 1MB of 6MB L3 cache per die in “Magny-Cours”
systems way 0 way 1 way 15 L3 cache Dir set
Implementation in fast SRAM (L3) minimizes – Access latency – Port occupancy of read-modify-write operations – Indirection latency for cache-to-cache transfers
56 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Die Block Diagram
“Magny-Cours” Die (Node)
Core Core Core Core Core Core 0 1 2 3 4 5
512kB 512kB 512kB 512kB 512kB 512kB L2 L2 L2 L2 L2 L2
System Request Interface (SRI)
L3 data array L3 tag (6MB)
XBAR
Memory Probe Controller Filter MCT/DCT
DRAM DRAM 4 HyperTransport™3 Ports DRAMDRAM DRAMDRAM
57 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Format of a Probe Filter Entry
. 16 probe filter entries per L3 cache line (64B), 4B per entry, 4-way set associative . 1MB of a 6MB L3 cache per die holds 256k probe filter entries and covers 16MB of cache
4 ways
Entry 0 Entry 1 Entry 2 Entry 3 L3 Cache Line (64B) 4 sets Entry 12 Entry 13 Entry 14 Entry 15
Probe Filter Entry (4B) Tag State Owner
EM, O, S, S1 states
58 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Probe Filter Transaction Scenarios
PF Hit PF Miss (*)
I O S S1 EM I O S S1 EM FETCH - D - - D - B B DI DI LOAD - D - - D - B B DI DI STORE - B B B DI - B B DI DI
Legend - Filtered “Effective” D Directed DI Directed Invalidate B Broadcast Invalidate “Ineffective” (*) PF miss implies line is Uncached (no broadcast necessary). State refers to the state of the line to be replaced upon allocation of new PF entry.
Traditional “Cache Hit Ratio” does not measure effectiveness of probe filter
59 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Probe Filter Coverage Ratio
Dir 0 5MB L3 + 3MB L2 5MB L3+ 3MB L2 Dir 2 Memory 0 Memory 3 256k lines 128k lines 128k lines 256k lines
P0 P2
Typical Worst case (Hotspotting) Uniformly distributed data Home node of each cached line is P0 Coverage ratio = 256k :: 128k = 2.0x Coverage ratio = 256k :: 128k * 4 = 0.5x
With sharing, a PF entry may P1 P3 track multiple cached copies and the coverage ratio increases
Dir 1 5MB L3 + 3MB L2 5MB L3 + 3MBL2 Dir 3 Memory 1 Memory 3 256k lines 128k lines 128k lines 256k lines
2 Socket “Magny-Cours”
60 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Cache Coherence Protocol
. Track lines in M, E, O or S state in probe filter
. PF is fully inclusive of all cached data in system - if a line is cached, then a PF entry must exist.
. Presence of probe filter entry says line in M, E, O or S state
. Absence of probe filter entry says line is uncached
. New messages – Directed probe on probe filter hit – Replacement notification E ->I (clean VicBlk)
61 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD Opteron™ EE Processor Improves Performance by up to 33%
Integer Throughput Performance 180 160 140 158 Up to 33% higher 120 performance with 100 119 than servers using
80 Score 60 40W ACP Quad-Core Higher is Better is Higher 40 2377 EE 2419 EE AMD Opteron™ 20 4 cores 6 cores 40W ACP 40W ACP processors 0 SPECint®_rate2006 – Two-Socket Servers
Low-Power Quad-Core AMD Opteron™ Processor (“Shanghai”) Model 2377 EE (2.3GHz/40W ACP) Low-Power Six-Core AMD Opteron™ Processor (“Istanbul”) Model 2419 EE (1.8GHz/40W ACP)
SPEC and SPECint are registered trademarks of the Standard Performance Evaluation Corporation. The result for Six-Core AMD Opteron™ processor Model 2419 EE is based upon data submitted to Standard Performance Evaluation Corporation as of August 5, 2009. The other results stated above reflect results published on http://www.spec.org/ as of August 5, 2009. The comparison presented above is based on the best performing two-socket servers using AMD Opteron™ processor Models 2419 EE and 2377 EE. For the latest results, visit www.spec.org. Please see backup slides for configuration information.
62 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 AMD Opteron™ EE Processor Improves Performance-per-Watt by up to 31%
Relative Performance to Power Ratios Workload: Server-side java/SPECpower®ssj_2008 150% Up to 31% higher 125% 131% performance-per- 100% watt than servers 100% 75% using 2.7GHz 75W 2384 2419 EE 75W ACP 40W ACP
50% ACP Quad-Core Higher is Better is Higher 1166 (overall 1614 (overall AMD Opteron™ 25% ssj_ops/watt) ssj_ops/watt) 412840 ssj_ops & 406954 ssj_ops & Percentage Improvement Percentage 260W @ 100% load 178W @ 100% load processors 0% point point Quad-Core AMD Opteron™ Processor (“Shanghai”) Model 2384 (2.7GHz/75W ACP) Low-Power Six-Core AMD Opteron™ Processor (“Istanbul”) Model 2419 EE (1.8GHz/40W ACP)
SPEC and SPECpower are registered trademarks of the Standard Performance Evaluation Corporation. The result for Six-Core AMD Opteron™ processor Model 2419 EE is based upon data submitted to Standard Performance Evaluation Corporation as of August 12, 2009. The other results stated above reflect results published on http://www.spec.org/ as of August 12, 2009. The comparison presented above is based on the best performing two-socket servers using AMD Opteron™ processor Models 2419 EE and 2384. For the latest results, visit www.spec.org. Please see backup slide 15 for configuration information.
63 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Density Calculation Details (Single Rack Comparison)
Notes
Six-Core AMD Opteron™ Six-Core AMD Opteron™ Six-Core AMD Opteron™ processor 2.6GHz processor Model 2425 HE processor Model 2419 EE Processor (75W ACP) (55W ACP) (40W ACP)
42U Standard rack 7.5KW power budget
Estimated KW power Estimated server power required to operate 42 ~13KW or ~ 9.2KW or ~7.5KW or consumption for a fully populated servers (1 rack) @ 42U rack (42 servers) 100% load point with (12894 watts) (9156 watts) (7476 watts) including 5% margin Platform Power 307W 218W 178W Consumption measured (100% load point / (100% load point / (100% load point / at 100% load point 550755 ssj_ops) 462215 ssj_ops) 406954 ssj_ops) Estimated number of 1U server units supported in a standard 24 servers 34 servers 42 servers 42U rack in power budget $305.55 $249.4 (up to 29% less than AMD (up to 40% less than AMD Estimated electricity cost per year. Annual Energy Costs $430.29 Opteron™ processor Model Opteron™ processor Assumes $.10/kWh and 1.6 PUE DC estimates* 2435 based server) Model 2435 based server) cooling factor
*For configuration and power consumption calculations see slide 3. SPEC and the benchmark name SPECpower_ssj are trademarks of the Standard Performance Evaluation Corporation. For the latest SPECpower_ssj2008 benchmark results, visit http://www.spec.org/power_ssj2008.
64 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Configuration Information (energy calc details*)
SIX-CORE SERVER SIX-CORE SERVER SIX-CORE SERVER ZT Systems 1224Ra ZT Systems 1224Ra ZT Systems 1224Ra 2x AMD Opteron™ processor 2x AMD Opteron™ processor 2x AMD Opteron™ processor model 2465 model 2425 HE model 2419 EE 16GB memory (4x 4048 MB) 16GB memory (4x 2048 MB) 16GB memory (4x 2048 MB) 1 SDD 1 SDD 1 SDD
* For full disclosure report, contact AMD. SPEC and the benchmark name SPECpower_ssj are trademarks of the Standard Performance Evaluation Corporation. For the latest SPECpower_ssj2008 benchmark results, visit http://www.spec.org/power_ssj2008.
65 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Configuration Information (EE vs. 75W ACP)
SIX-CORE SERVER QUAD-CORE SERVER ZT Systems 1224Ra ZT Systems 1224Ra 2x AMD Opteron™ processor 2x AMD Opteron™ processor Model 2419 EE (55W ACP) Model 2384 75W ACP 16GB memory (4x 2048 MB) 16GB memory (4x 2048 MB) 1 SDD 1 SDD
* For full disclosure report, contact AMD. SPEC and the benchmark name SPECpower_ssj are trademarks of the Standard Performance Evaluation Corporation. For the latest SPECpower_ssj2008 benchmark results, visit http://www.spec.org/power_ssj2008.
66 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Footnotes
Two-Socket SPECint®_rate2006 158 using 2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2419 EE in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 250GB SATA disk drive, Red Hat Enterprise Linux® Server 5.2 64-bit 119 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2377 EE in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 300GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP1 64-bit http://www.spec.org/cpu2006/results/res2009q2/cpu2006-20090406-06935.html
67 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HT Assist • HT Assist helps reduces memory latency and increase overall system performance in multi-socket systems • HT Assist improves HyperTransport™ technology link efficiency and increases performance by: • Reducing probe traffic • Resolving probes more quickly • Probe “broadcasting” can be eliminated in 8 of 11 typical CPU-to-CPU transactions • HT Assist can provide a significant benefit in 4-way and greater systems • 4-Way Stream memory bandwidth performance improves by ~60% (41.5GB/s with HT Assist vs. 25.5GB/s without HT Assist) • HT Assist reserves 1MB portion of each CPU’s L3 cache to act as a directory. This directory tracks where that CPU’s cache lines are used elsewhere in the system • For 2-way systems this reduction in L3 cache size may eliminate any HT Assist benefit as probe traffic is already significantly less than in 4-way systems • Each CPU is considered the “host” of the cache information contained in its L3 directory • For many CPU-to-CPU transactions the host CPU knows exactly which CPU to probe for the information it needs, eliminating the need to “broadcast”. • Reducing broadcasting cuts down considerably the amount of system probe traffic and helps resolve probes more quickly, resulting in reduced memory latency and improved system performance
68 AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
6 For more information be sure to visit AMD at SC’09 booth #1417 8 Two-Socket SPECint®_rate2006 – Slide 8 & 11
205 using 2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2435 in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 250GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit
136 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2384 in Supermicro A+ Server 1021M- UR+B server, 32GB (8x4GB DDR2-800) memory, 250GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081024-05683.html
217 using 2 x Intel Xeon processors (“Gainestown”) Model E5540 in Supermicro SuperServer 6026T-NTR+ server, 24GB (12x2GB DDR3-1066) memory, 150GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2009q1/cpu2006-20090316-06749.html
131 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2382 in Dell PowerEdge R805 server, 32GB (8x4GB DDR2-800) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081110-05955.html
51.5 using 2 x Dual-Core AMD Opteron™ processors Model 2218 in Fujitsu Siemens Computers PRIMERGY RX330 S1 server, 32GB (8x4GB DDR2-667) memory, 36GB SAS disk drive, SuSE Linux® Enterprise Server 10 64-bit http://www.spec.org/cpu2006/results/res2007q2/cpu2006-20070427-00947.html
139 using 2 x Quad-Core Intel Xeon processors (“Harpertown”) Model E5450 in Dell PowerEdge 2950 III server, 16GB (4x4GB DDR2-667 FB-DIMM) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081027-05790.html
69 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Two-Socket SPECfp®_rate2006 - Slide 8 & 11
143 using 2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2435 in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 250GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit
118 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2384 in Supermicro A+ Server 1021M- UR+B server, 32GB (8x4GB DDR2-800) memory, 250GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081024-05684.html
115 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2382 in Dell PowerEdge R805 server, 32GB (8x4GB DDR2-800) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081110-05954.html
48.5 using 2 x Dual-Core AMD Opteron™ processors Model 2218 in Fujitsu Siemens Computers PRIMERGY RX330 S1 server, 32GB (8x4GB DDR2-667) memory, 36GB SAS disk drive, SuSE Linux® Enterprise Server 10 64-bit http://www.spec.org/cpu2006/results/res2007q2/cpu2006-20070427-00945.html
169 using 2 x Intel Xeon processors (“Gainestown”) Model E5540 in Supermicro SuperServer 6026T-NTR+ server, 24GB (12x2GB DDR3-1066) memory, 150GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2009q1/cpu2006-20090316-06747.html
78.4 using 2 x Quad-Core Intel Xeon processors (“Harpertown”) Model E5450 in Dell PowerEdge 2950 III server, 16GB (4x4GB DDR2-667 FB-DIMM) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081027-05795.html
70 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Two-Socket SPECjbb®2005 – Slide 8 & 11
462189 score using 2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2435 in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 300GB SATA disk drive, Microsoft® Windows Server® 2003 Enterprise x64 Edition SP2 352700 score using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2384 in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 300GB SATA disk drive, Microsoft® Windows Server® 2003 Enterprise x64 Edition SP2 http://www.spec.org/osg/jbb2005/results/res2008q4/jbb2005-20081024-00551.html 468132 score using 2 x Intel Xeon processors (“Gainestown”) Model E5540 in Supermicro X8DTN+ motherboard, 12GB (6x2GB DDR3-1066) memory, 150GB SATA disk drive, Microsoft® Windows Server® 2008 R2 Enterprise x64 Edition SP1 362239 using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 2382 in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 300GB SATA disk drive, Microsoft® Windows Server® 2008 Enterprise x64 Edition 66420 using 2 x Dual-Core AMD Opteron™ processors Model 2218 in Tyan Thunder K9HM (S3992) server, 8GB (16x512MB DDR2-667) memory, 120GB IDE disk drive, Microsoft® Windows Server® 2003 Enterprise Edition SP1 64-bit http://www.spec.org/osg/jbb2005/results/res2006q3/jbb2005-20060718-00154.html
310028 using 2 x Quad-Core Intel Xeon processors (“Harpertown”) Model E5450 in IBM BladeCenter HS21XM server, 16GB (8x2GB DDR2-667) memory, 73GB SAS disk drive, Microsoft® Windows Server® 2003 R2 Enterprise x64 Edition SP1 http://www.spec.org/osg/jbb2005/results/res2008q2/jbb2005-20080527-00494.html
71 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Two-Socket STREAM – Slide 7 & 8
21GB/s using 2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2435 in Supermicro H8DMU+ motherboard, 16GB (8x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit 21GB/s using 2 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in Supermicro H8DMU+ motherboard, 16GB (8x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit
Two-Socket Server Idle Power – Slide 11 83.21W using 2 x Dual-Core AMD Opteron™ processors Model 2218 in ZT Systems 1224Ri Datacenter Server, 16GB (4x4GB DDR2-800) memory, 80GB SSD SATA disk drive, Sparkle Power International LTD SPI4001UG power supply, Microsoft® Windows Server® 2008 SP1 64-bit
92.84W using 2 x Quad-Core AMD Opteron™ (“Shanghai”) processors Model 2218 in ZT Systems 1224Ri Datacenter Server, 16GB (4x4GB DDR2-800) memory, 80GB SSD SATA disk drive, Sparkle Power International LTD SPI4001UG power supply, Microsoft® Windows Server® 2008 SP1 64-bit
91.32W using 2 x Six-Core AMD Opteron™ (“Istanbul”) processors Model 2218 in ZT Systems 1224Ri Datacenter Server, 16GB (4x4GB DDR2-800) memory, 80GB SSD SATA disk drive, Sparkle Power International LTD SPI4001UG power supply, Microsoft® Windows Server® 2008 SP1 64-bit
72 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Four-Socket and Eight-Socket SPECint®_rate2006 – Slide 6 & 8
402 using 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Thunder n4250QE server (S4985-SI), 64GB (16x4GB DDR2-800) memory, 250GB SATA disk drive, Red Hat Enterprise Linux® Server release 5.3 64-bit http://www.spec.org/cpu2006/results/res2009q2/cpu2006-20090511-07355.html
249 using 4 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in HP ProLiant DL585 G5 server, 64GB (16x4GB DDR2-800) memory, 146GB SAS disk drive, Red Hat Enterprise Linux® Server release 5.2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081027-05740.html
386 using 8 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in Sun Fire X4600 M2 server, 128GB (64x2GB DDR2-667) memory, 2x 72GB SAS disk drive, OpenSolaris 2008.05 http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081208-06223.html
294 using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model X7460 in IBM System x3850 M2 server, 64GB (16x4GB DDR2-667 FB-DIMM) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20080915-05319.html
253 using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model E7450 in Intel Server System S7000FC4UR motherboard, 32GB (16x2GB DDR2-667 FB-DIMM) memory, 73GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20080915-05347.html
73 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Four-Socket SPECjbb®2005 – Slide 8
866261 using 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Transport TX46 server, 64GB (16x4GB DDR2-800) memory, 250GB SATA disk drive, Microsoft® Windows Server® 2008 Enterprise x64 Edition
721843 using 4 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in IBM Bladecenter LS42 server, 64GB (16x4GB DDR2-800) memory, 36GB SAS disk drive, Microsoft® Windows Server® 2008 Enterprise x64 Edition http://www.spec.org/osg/jbb2005/results/res2008q4/jbb2005-20081112-00559.html
633897 using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model X7460 in Fujitsu PRIMERGY RX600 S4 server, 64GB (16x4GB DDR2-667 FB-DIMM) memory, 36GB SAS disk drive, Microsoft® Windows Server® 2003 Enterprise x64 Edition http://www.spec.org/osg/jbb2005/results/res2009q1/jbb2005-20090305-00664.html
74 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Four-Socket SPECfp®_rate2006 – 8
276 using 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Transport TX46 server, 64GB (16x4GB DDR2-800) memory, 250GB SATA disk drive, Red Hat Enterprise Linux® Server release 5.2 64-bit
212 using 4 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in Dell PowerEdge M905 server, 64GB (16x4GB DDR2-800) memory, 36GB SAS + 73GB SAS disk drives, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20081110-05969.html 156 using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model X7460 in HP ProLiant DL580 G5 server, 64GB (16x4GB DDR2-667 FB-DIMM) memory, 146GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP1 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20080911-05253.html
139 using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model E7450 in Fujitsu Siemens Computers PRIMERGY RX600 S4 server, 64GB (16x4GB DDR2-667 FB-DIMM) memory, 36GB SAS disk drive, SuSE Linux® Enterprise Server 10 SP2 64-bit http://www.spec.org/cpu2006/results/res2008q4/cpu2006-20080922-05389.html
75 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Four-Socket STREAM – Slide 7 & 8
42GB/s using 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Thunder n4250QE (S4985-E) motherboard, 32GB (16x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP1 64- bit (with HT Assist enabled) 25.5GB/s using 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Thunder n4250QE (S4985-E) motherboard, 32GB (16x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit (with HT Assist disabled) 24GB/s using 4 x Quad-Core AMD Opteron™ processors (“Shanghai”) Model 8384 in Tyan Thunder n4250QE (S4985-E) motherboard, 32GB (16x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit 9GB/s using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model X7460 in Supermicro X7QC3+ motherboard, 32GB (16x2GB DDR2-667 FB-DIMM) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit 9GB/s using 4 x Hex-Core Intel Xeon processors (“Dunnington”) Model E7450 in Supermicro X7QC3+ motherboard, 32GB (16x2GB DDR2-667 FB-DIMM) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit
76 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 2 Socket HPL Performance – Slide 10
AMD system:
2 x Six-Core AMD Opteron™ processors ("Istanbul") Model 2435 (2.6GHz) in Supermicro A+ Server 1021M-UR+B server, 32GB (8x4GB DDR2-800) memory, 300GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP1
64-bit Intel system:
2 x Quad-Core Intel processor ("Nehalem") Model X5570 (2.93 GHz) 2in Intel board S5520UR, 24GB (6x4GB DDR3-1333) memory, 300GB SATA disk drive, SuSE Linux® Enterprise Server 10 SP1 64-bit
77 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 HPC Interconnect performance testing – slide 18
Configurations used to collect performance data:
• ausqc[01-08]: 2 x Quad-Core AMD Opteron™ processors Model 2384 in Supermicro H8DMU+ motherboard, 16GB (8x2GB DDR2-800 ), 150GB IDE disk drive, IB DDR [card,switch], SuSE Linux® Enterprise Server 10 SP1 64-bit • fiorano[01-08]: 2 x Quad-Core AMD Opteron™ processors Model 2387 in Toonie 2 , 16GB (8x2GB DDR2-800 ), 150GB IDE disk drive, IB QDR [card,switch], SuSE Linux® Enterprise Server 10 SP2 64-bit • Tooniehpc[1,2]: 2 x Quad-Core AMD Opteron™ processors Model 2387 in Toonie 2 , 16GB (8x2GB DDR2-800 ), 150GB IDE disk drive, IB [D,QDR] back to back [no IB switch], SuSE Linux® Enterprise Server 10 SP2 64-bit • IB [D,Q]DR cards and switches provided by Mellanox.
78 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 STREAM (Memory Bandwidth) – From slide 31
2 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 2435 in Supermicro A+ Server 1021M-UR+B server, 16GB (8x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 4 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435 in Tyan Transport TX46 server, 32GB (16x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 8 x Six-Core AMD Opteron™ processors (“Istanbul”) Model 8435, AMD 5690 Chipset reference design platform, 64GB (32x2GB DDR2-800) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 2 x AMD Opteron™ processors (“Magny-Cours ”) Model 61xx, AMD 5690 Chipset reference design platform, 32GB memory, SuSE Linux® Enterprise Server 10 SP2 64-bit. 4 x AMD Opteron™ processors (“Magny-Cours”) Model 61xx, AMD 5690 Chipset reference design platform, 64GB memory, SuSE Linux® Enterprise Server 10 SP2 64-bit. 2x Intel Xeon processors (“Gainestown”) Model X5570 in Supermicro SuperServer 6026T-NTR+ server, 24GB memory (6x4GB DDR3-1333) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 2x Intel Xeon processors (“Gainestown”) Model E5540 in Supermicro SuperServer 6026T-NTR+ server, 24GB memory (6x4GB DDR3-1066) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 2x Intel Xeon processors (“Gainestown”) Model L5520 in Supermicro SuperServer 6026T-NTR+ server, 24GB memory (6x4GB DDR3-1066) memory, SuSE Linux® Enterprise Server 10 SP2 64-bit 4x Intel Xeon processors (“Dunnington”) Model X7460 in Supermicro X7QC3+ motherboard, 32GB (16x2GB FBDIMM) memory, SuSE Linux® Enterprise Server 10 SP1 64-bit
79 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Disclaimer & Attribution
DISCLAIMER The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors.
The information contained herein is subject to change and may be rendered inaccurate for many reasons, including but not limited to product and roadmap changes, component and motherboard version changes, new model and/or product releases, product differences between differing manufacturers, software changes, BIOS flashes, firmware upgrades, or the like. AMD assumes no obligation to update or otherwise correct or revise this information. However, AMD reserves the right to revise this information and to make changes from time to time to the content hereof without obligation of AMD to notify any person of such revisions or changes.
AMD MAKES NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE CONTENTS HEREOF AND ASSUMES NO RESPONSIBILITY FOR ANY INACCURACIES, ERRORS OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION.
AMD SPECIFICALLY DISCLAIMS ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE. IN NO EVENT WILL AMD BE LIABLE TO ANY PERSON FOR ANY DIRECT, INDIRECT, SPECIAL OR OTHER CONSEQUENTIAL DAMAGES ARISING FROM THE USE OF ANY INFORMATION CONTAINED HEREIN, EVEN IF AMD IS EXPRESSLY ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
ATTRIBUTION © 2009 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, ATI, the ATI logo, AMD CoolCore, AMD PowerNow!, AMD Opteron, AMD Virtualization, AMD-V, Dual Dynamic Power Management, Catalyst, FireGL, FirePro, FireStream, Radeon, and combinations thereof are trademarks of Advanced Micro Devices, Inc. HyperTransport is a licensed trademark of the HyperTransport Technology consortium. Microsoft, Windows, and DirectX are registered trademarks of Microsoft Corporation in the United States and/or other jurisdictions. Other names are for informational purposes only and may be trademarks of their respective owners.
OpenCL is a trademark of Apple Inc. used under License to the Khronos Group Inc.
80 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417 Index Slide
•HPC Product Portfolio •Platform Roadmap •Istanbul •Istanbul Developer Update •CPU overview through 2010 •Magny-Cours •A peek at the future •STREAM Computing •EyeFinity Overview •Backup
81 | AMD HPC Product Portfolio Update @ SC’09 | November 30, 2009
For more information be sure to visit AMD at SC’09 booth #1417